Suraphan Thawornwong David Enke Cihan Dagli

Stock trend or stock price prediction is an important financial subject that has attracted researchers’ attention for many years. This is due to the fact that a successful prediction model could result in substantial monetary rewards. However, predicting stock price or stock return is not a simple task, since many market factors are involved and their structural relationships are too complex to be clearly extracted. Technical analysis has a long tradition in forecasting movements in a financial time series (Plummer 1989); however, it also has a long history of being criticized by academics and practitioners (Malkiel 1995).

This criticism can be explained by the fact that technical analysis is built on weak foundations. For example, the expectation that some historical patterns of a stock price will be repeated in the future may not necessarily be fulfilled, since market conditions change over time (Wong and Ng 1994) and there is no explanation as to why these patterns should be expected to repeat (Jegadeesh 2000). As noted in the efficient market hypothesis (Fama 1970), it is impossible to forecast future price, since the price already reflects everything that is currently known about the stock.

Nonetheless, in recent years technical analysis has been widely accepted as one of many viable analytical

options among both financial economists and brokerage firms (Achelis 1995). In fact, large investments are rarely made without touching this technical environment. This is due to the fact that many researchers are open to the idea that markets may not be fully efficient and prices may be affected by human sentiments found in psychology rather than economics (Barberis et al. 1998).

Technical analysis also appears to be a compromising tool, since it offers a relative mixture of human, political, and economical events. Theoretically, it attempts to predict the trend of stock prices by using data surrounding past prices and volumes. The main problem with this technique is that it relies heavily on the discovery of strong empirical regularities in observations of the price and volume movements (Liu and Lee 1997). In other words, the supporter of this technique is only concerned with the identification of major turning points for assessing the securities’ movement. In reality, these regularities are not always evident, often masked by noise, and vary from security to security, making it difficult for investors who use this technique to consistently and accurately determine future prices.

In addition to the increased use of technical analysis, investors today are becoming more dependent on advanced computer algorithms and technologies to benefit from a wider range of investment choices (Elton and Gruber 1991). Artificial neural networks are one of the technologies that have caused the most excitement in this financial environment. They provide an

313

interesting technique that can theoretically approximate any nonlinear continuous function on a compact domain to any designed degree of accuracy (Cybenko 1989). The novelty of neural networks lies in the ability to model nonlinear processes without a priori assumption about the nature of the generating process (Hagen et al. 1996). This is useful in security investment and other financial areas